Systems and methods for magnetic resonance imaging

Abstract

A method for magnetic resonance imaging (MRI) may include obtaining a magnetic resonance (MR) image of a subject, wherein the MR image may be acquired based on a first MRI device and include at least one region of interest (ROI) of the subject. The method may also include selecting, based on the MR image and an ROI determination model, a portion of a main magnetic field generated by the first MRI device. The selected portion of the main magnetic field may correspond to the at least one ROI. The method may also include performing a magnetic field homogenization operation on the selected portion of the main magnetic field.

What is claimed is: 
     
       1. A system for magnetic resonance imaging (MRI), comprising:
 at least one storage device storing a set of instructions; and 
 at least one processor in communication with the at least one storage device, when executing the stored set of instructions, the at least one processor causes the system to perform operations including:
 obtaining a magnetic resonance (MR) image of a subject, wherein the MR image is acquired based on a first MRI device and includes at least one region of interest (ROI) of the subject; 
 selecting, based on the MR image and an ROI determination model, a portion of a main magnetic field generated by the first MRI device, the selected portion of the main magnetic field corresponding to the at least one ROI; and 
 performing a magnetic field homogenization operation on the selected portion of the main magnetic field. 
 
 
     
     
       2. The system of  claim 1 , wherein the selecting, based on the MR image and the ROI determination model, a portion of the main magnetic field generated by the first MRI device includes:
 generating, based on the ROI determination model and the MR image, at least one mask image, one of the at least one mask image corresponding to one or more of the at least one ROI; 
 obtaining a magnetic field map of the subject, the magnetic field map including the at least one ROI; 
 obtaining at least one ROI image by segmenting the at least one ROI from the magnetic field map based on the at least one mask image; and 
 selecting, based on the at least one ROI image, the portion of the main magnetic field on which the field homogenization operation is performed. 
 
     
     
       3. The system of  claim 2 , wherein the ROI determination model is obtained according to a first training process including:
 obtaining a plurality of groups of first training samples; and 
 generating the ROI determination model by training a first preliminary model using the plurality of groups of first training samples. 
 
     
     
       4. The system of  claim 3 , wherein the generating the ROI determination model by training the first preliminary model using the plurality of groups of first training samples includes:
 initializing the first preliminary model; 
 generating the ROI determination model by updating the initialized first preliminary model using a first iteration process including a plurality of iterations, each of the plurality of iterations of the first iteration process including:
 obtaining one of the plurality of groups of first training samples that includes a first sample input image and at least one corresponding reference mask image relating to at least one reference ROI of the first sample input image; 
 generating at least one intermediate mask image by inputting the first sample input image of the group of first training sample into a first intermediate model, the first intermediate model being the initialized first preliminary model in a first iteration of the plurality of iterations of the first iteration process or a previously updated model generated in a previous iteration in the first iteration process, the intermediate mask image including at least one candidate ROI of the first sample input image; 
 determining a value of a first cost function based on the at least one intermediate mask image and the at least one reference mask image of the group of first training sample; 
 determining whether a first termination condition is satisfied; 
 in response to determining that the first termination condition is not satisfied,
 generating an updated model by updating the first intermediate model; and 
 initiating a next iteration; and 
 
 
 determining the updated model generated in a last iteration of the plurality of iterations of the first iteration process as the ROI determination model. 
 
     
     
       5. The system of  claim 4 , wherein the value of the first cost function is determined based on at least one of
 a difference between a size of the at least one candidate ROI in the at least one intermediate mask image and a size of the at least one reference ROI in the at least one reference mask image, or 
 a difference between a location of the at least one candidate ROI in the at least one intermediate mask image and a location of the at least one reference ROI in the at least one reference mask image. 
 
     
     
       6. The system of  claim 4 , wherein the first termination condition relates to at least one of the value of the first cost function or a count of iterations of the first iteration process that have been performed. 
     
     
       7. The system of  claim 2 , wherein the generating, based on the ROI determination model and the MR image, the at least one mask image includes:
 preprocessing the MR image; and 
 generating the at least one mask image based on the ROI determination model and the preprocessed MR image. 
 
     
     
       8. The system of  claim 7 , wherein the preprocessing the MR image includes at least one of the following operations:
 performing a phase unwrapping operation on the MR image; or 
 preprocessing the MR image based on a preprocessing model, the preprocessed MR image having a higher image quality than the MR image. 
 
     
     
       9. The system of  claim 8 , wherein the image quality relates to an image resolution. 
     
     
       10. The system of  claim 8 , wherein the preprocessing model is obtained according to a second training process including:
 obtaining a plurality of groups of second training samples; and 
 generating the preprocessing model by training a second preliminary model using the plurality of groups of second training samples. 
 
     
     
       11. The system of  claim 10 , wherein the generating the preprocessing model by training the second preliminary model using the plurality of groups of second training samples includes:
 initializing the second preliminary model; 
 generating the preprocessing model by updating the initialized second preliminary model using a second iteration process including a plurality of iterations, each of the plurality of iterations of the second iteration process including:
 obtaining one of the plurality of groups of second training samples that includes a second sample input image and a corresponding reference image, the second sample input image having a higher image quality than the corresponding reference image; 
 generating an intermediate image by inputting the second sample input image of the group of first training sample into a second intermediate model, the second intermediate model being the initialized second preliminary model in a first iteration of the plurality of iterations of the second iteration process or a previously updated model generated in a previous iteration in the second iteration process; 
 determining a value of a second cost function based on the intermediate image and the reference image of the group of second training sample; 
 determining whether a second termination condition is satisfied; 
 in response to determining that the second termination condition is not satisfied,
 generating an updated model by updating the second intermediate model; and 
 initiating a next iteration; and 
 
 
 determining the updated model generated in a last iteration of the plurality of iterations of the second iteration process as the preprocessing model. 
 
     
     
       12. The system of  claim 11 , wherein the second sample input image and the corresponding reference image of at least one of the plurality of groups of second training samples are obtained by scanning a sample subject using a second MRI device. 
     
     
       13. The system of  claim 11 , wherein
 the reference image of at least one of the plurality of groups of second training samples is obtained by scanning a sample subject using a third MRI device, and 
 the corresponding second sample input image of the at least one of the plurality of groups of second training samples is obtained by processing the reference image. 
 
     
     
       14. The system of  claim 11 , wherein the value of the second cost function is determined based on at least one of
 a difference between pixel values of pixels of the intermediate image and pixel values of pixels of the reference image, or 
 a difference between a homogenization degree of the main magnetic field determined based on the intermediate image and a degree threshold. 
 
     
     
       15. The system of  claim 11 , wherein the second termination condition relates to the value of the second cost function, or a count of iterations of the second iteration process that have been performed. 
     
     
       16. The system of  claim 8 , wherein the preprocessing model is constructed based on at least one of a U-shape network (U-Net), a generative adversarial network (GAN), or a recurrent generative adversarial network. 
     
     
       17. The system of  claim 1 , wherein the ROI determination model is constructed based on a U-shape network (U-Net). 
     
     
       18. The system of  claim 1 , wherein the magnetic field homogenization operation is performed on the selected region of the main magnetic field based on at least one homogeneity threshold each of which corresponds to one of the at least one ROI. 
     
     
       19. A system for magnetic resonance imaging (MRI), comprising:
 at least one storage device storing a set of instructions; and 
 at least one processor in communication with the at least one storage device, when executing the stored set of instructions, the at least one processor causes the system to perform operations including:
 obtaining a magnetic resonance (MR) image of a subject, wherein the MR image includes at least one region of interest (ROI) of the subject; 
 preprocessing the MR image based on a preprocessing model, the preprocessed MR image having a higher image quality than the MR image; 
 generating at least one mask image based on an ROI determination model and the preprocessed MR image, one of the at least one mask image corresponding to one or more of the at least one ROI; and 
 obtaining at least one ROI image based on the at least one mask image. 
 
 
     
     
       20. A method for magnetic resonance imaging (MRI) implemented on a machine having at least one processor and at least one storage device, comprising:
 obtaining a magnetic resonance (MR) image of a subject, wherein the MR image is acquired based on a first MRI device and includes at least one region of interest (ROI) of the subject; 
 selecting, based on the MR image and an ROI determination model, a portion of a main magnetic field generated by the first MRI device, the selected portion of the main magnetic field corresponding to the at least one ROI; and 
 performing a magnetic field homogenization operation on the selected portion of the main magnetic field.